Is there a way to create a torch.Size directly?

How I workaround it currently is create a dummy tensor with the dimensions i want and call .size() on it. Seems kinda wrong.

The reason I want torch.Size is cause I was making my own dataset class and wanted a sliding window batching kind of operation on it and thought it would be easiest to set_() the shape and strides of a view of the data.

torch.Size([1, 2, 3, 4])